Do generalist CEOs engage in more tax avoidance than specialist CEOs?

Abstract

Existing research suggests that generalist CEOs, who possess managerial skills that are transferrable across firms and industries, are more able to bear downside risk than specialist CEOs with non-transferrable managerial expertise. In this paper, we examine whether generalist CEOs are better positioned to engage in risky tax avoidance strategies than specialist CEOs. Our empirical results support this prediction and show that firms with generalist CEOs tend to engage in more tax avoidance than those with specialist CEOs. Our identification strategy includes an instrumental variable method and a difference-in-differences test using CEO turnover as a quasi-natural experiment to correct for endogeneities. A battery of robustness checks and cross-sectional tests strengthens our findings. Taken together, our findings imply that general managerial skills of CEOs matter more for tax planning than do specific managerial skills.

Keywords:

• Tax avoidance
• Tax uncertainty
• CEO skills
• Agency problem

Disclosure statement

No potential conflict of interest was reported by the author(s).

Acknowledgement

We would like to thank Martin Jacob (Associate Editor), two anonymous referees, Mark Anderson, Hussein Warsame, Abu Rahaman, Philip Beaulieu, Alan Macnaughton, Alex Edwards (discussant), Rahat Jafri, Mahfuz Chy and the participants at the 2018 Ph.D. Colloquium of Canadian Academic Accounting Association and the 2018 Financial Reporting and Business Communication conference at University of Bristol, U.K for their helpful comments and suggestions on the earlier version of the paper. Harun Rashid would like to thank Haskayne School of Business, University of Calgary and College of Business Administration and Public Policy, California State University, Dominguez Hills for financial support. Muhammad Kabir would like to thank the Asper School of Business for financial support.

Notes

1 We use the term “tax avoidance” to refer to a strategy of reducing explicit taxes or tax liabilities. A tax avoidance strategy may involve real firm activities that reduce taxes and/or lobbying activities for tax benefits (Hanlon and Heitzman, 2010). Consistent with Dyreng et al. (2008), our definition of ‘tax avoidance’ measure is broad and may include low risk tax planning activities, such as investment in municipal bonds, to highly risky tax planning such as participating in tax shelters. In this paper, tax avoidance and tax planning are used interchangeably.

2 We acknowledge that the experience of generalist CEOs is not likely to be fully exogenous, and be reflective of other personal characteristics, such as risk aversion or risk-taking. For example, certain risk-prone individuals may be more likely to self-select into a career track that is likely to yield a generalist classification. We thank an anonymous reviewer for this insight.

3 Prior studies also find that managerial equity risk incentives are associated with other corporate policy choices such as investing and financing (Guay, 1999, Rajgopal and Shevlin, 2002, Coles et al., 2006, Cohen et al., 2009).

4 We thank Custodio et al. (2013) for sharing GAI data.

5 We thank Kim and Lu (2011) for sharing their CEO ownership data.

6 We acknowledge that CEO turnovers may not be strictly exogenous, and the GAI of the successor may not be completely random or exogenous. For instance, CEO turnover may take place for paying too much or too little taxes due to reputational concerns (Chyz and Gaertner, 2018). Having said that, we find that the bulk of our results support the view that a link exists between a generalist CEO and tax avoidance. We thank an anonymous reviewer for this comment.

7 As we control for firm and year fixed effects, the main effect of TREATED and post are subsumed and not reported in the table (Amiram et al., 2019). We do not control for the firm-CEO fixed effect, since we do not allow the status of a generalist or specialist CEO to change in our DiD sample.

8 As the DiD sample is a small subsample of the full sample, we use a parsimonious model and do not include ΔNOL and Capex to maximize the DiD sample size.

9 A possible concern is that instead of representing generalist CEO skills, the ‘generalist CEO’ dummy variable may actually be a proxy for company outsider CEO. To mitigate such concerns, we control for company outsider CEO replacement. We find that out of 95 CEO replacements in our sample, 18 replacements represent company outsider CEOs of which 7 are specialist CEOs and 11 are generalist CEOs, which is 18.95% outsider CEO replacements. This is comparable to Huson et al. (2001). We find (untabulated) that, even after controlling for industry outsider CEO replacements, the coefficient of TREATED*POST is significant at the 10% level.

10 PRE and POST dummy variables for before and after CEO appointment years capture whether or not CEOs’ appointments have any effect on tax planning regardless of their generalist or specialist status.

11 One could argue that we are simply capturing the effect of CEO replacement on tax avoidance instead of the effect of general skills of CEOs. In such a case, we would observe similar effects on tax avoidance when a generalist CEO is replaced by a specialist CEO. To address this concern, we run an alternative DiD test where the pseudo-treated firms are those where generalist CEOs are replaced by specialist CEOs. We do not find any significant evidence that firms’ tax avoidance increases after generalist CEOs are replaced by specialist CEOs. In the Internet appendix, we present detailed arguments for this point and report the results in Table SA 3.

12 Using the full sample, we also perform OLS regressions, where GAI is the explanatory variable of interest and Cash ETR and the three-year average Cash ETR are dependent variables. We find that our results are qualitatively and quantitatively significant. We present the results in the Internet appendix in Table SA 1. In addition, as loss firms represent a significant percentage of the corporate population (Henry and Sansing, 2018), we perform regression including loss firms by replacing Cash ETR with cash-tax/asset and UTB measures. Our results remain qualitatively similar, as reported in the Internet appendix in Table SA 2.

13 A number of studies point out several limitations of GAAP ETR. For example, Drake et al. (2020) document that GAAP ETRs are likely to be downward biased because of changes in valuation allowance, which is not directly related to tax avoidance. Similarly, GAAP ETR is also affected by goodwill impairments, which have no relation to tax avoidance (King et al. 2020). Nonetheless, our results hold for GAAP ETR as a proxy for tax avoidance.

14 Companies are required to disclose UTB-related information when their tax positions are less than 50% likely to be sustained upon a tax audit.

15 To maximize the number of matchings, we allow four matches with replacement.

16 We acknowledge that while the matching technique mitigates selection bias to a great extent, it does not completely alleviate it due to its inability to control for unobservable characteristics (Lennox et al., 2012).

17 Our results are similar if we perform the matching estimate using two stages: create a matched sample in the first stage and in the second stage, run an OLS regression with the matched sample with control variables, firm-CEO, and year fixed effect.

18 A concern may arise that we are simply capturing the effect of change of an executive instead of the effect of generalist CEOs on tax planning. We believe that a CEO’s general skills are driving the results for the following reasons. First, as a generalist, the new CEO has a relatively larger network than a specialist’s network. This enables the CEO to search from a larger pool of candidates and appoint the best one as the CFO for better financial performance including tax planning. Second, had the result been completely driven simply by the executive or CEO changes instead of general skillsets, we would have a significant increase in tax avoidance with the new specialist CEO. Our alternative DiD test in the Internet appendix (Table SA 3) does not support this notion. In Table SA 3, we examine whether or not the change of generalist CEO by the specialist CEO increases tax avoidance and do not find any significant result. Last but not the least, we would not likely have a significant result for the mechanisms for risky tax avoidance for generalist CEOs. As shown by our empirical tests in Table 10, we find results supporting this prediction. If our results were due to CEO changes, it would be reasonable to expect that the magnitude of risky tax avoidance strategies would not vary with regards to the differences in the general skillsets as captured in the GAI.

19 The tax haven data is collected from Scott Dyreng’s website https://sites.google.com/site/scottdyreng/Home/data-and-code. We thank him for sharing the data.

20 As we use the noncompete enforcement index (NEI) as the instrumental variable for our baseline 2SLS regression, it is reasonable to use the same instrument for this cross-sectional test. However, NEI may affect intangibles such as trade secrets (Glaeser, 2018) and thus, cannot be a valid instrument when the Intangibles is the dependent variable. Therefore, following prior literature (Dyreng and Lindsey 2009, Koester et al. 2017), we use OLS and logit regressions for all the dependent variables in Table 10. However, we also run our baseline 2SLS model for Tax-Shelter and Tax-Havens and find qualitatively similar results (untabulated).